]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
13 years agoMFV of tzdata2011c, r219409
Edwin Groothuis [Tue, 8 Mar 2011 21:11:31 +0000 (21:11 +0000)]
MFV of tzdata2011c, r219409

- No leapsecond in June 2011
- Add and update timezones for America/Juneau, America/Sitka,
  America/Metlakatla
- Change of Summer time to Winter time in Chili happens in April
  2011, not in March 2011.

13 years agoVendor import of tzdata2011c
Edwin Groothuis [Tue, 8 Mar 2011 21:06:17 +0000 (21:06 +0000)]
Vendor import of tzdata2011c

- No leapsecond in June 2011
- Add and update timezones for America/Juneau, America/Sitka,
  America/Metlakatla
- Change of Summer time to Winter time in Chili happens in April
  2011, not in March 2011.

Obtained from: ftp://elsie.nci.nih.gov/pub/

13 years agoStop hard-coding default directory mode as 0777.
Jung-uk Kim [Tue, 8 Mar 2011 20:13:29 +0000 (20:13 +0000)]
Stop hard-coding default directory mode as 0777.

13 years agoRearrange dc_tx_underrun() a bit to correctly set TX FIFO threshold
Pyun YongHyeon [Tue, 8 Mar 2011 19:49:16 +0000 (19:49 +0000)]
Rearrange dc_tx_underrun() a bit to correctly set TX FIFO threshold
value. Controllers that always require "store and forward" mode(
Davicom and PNIC 82C168) have no way to recover from TX underrun
except completely reinitializing hardware. Previously only Davicom
was reinitialized and the TX FIFO threshold was changed not to use
"store and forward" mode after reinitialization since the default
FIFO threshold value was 0. This effectively disabled Davicom
controller's "store and forward" mode once it encountered TX
underruns. In theory, this can cause watchodg timeouts.

Intel 21143 controller requires TX MAC should be idle before
changing TX FIFO threshold. So driver tried to disable TX MAC and
checked whether it saw the idle state of TX MAC. Driver should
perform full hardware reinitialization on failing to enter to idle
state and it should not touch TX MAC again once it performed full
reinitialization.

While I'm here remove resetting TX FIFO threshold to 0 when
interface is put into down state. If driver ever encountered TX
underrun, it's likely to trigger TX underrun again whenever
interface is brought to up again. Keeping old/learned TX FIFO
threshold value shall reduce the chance of seeing TX underrns in
next run.

13 years agoBump __FreeBSD_version for struct sysvec (sv_schedtail) changes.
Dmitry Chagin [Tue, 8 Mar 2011 19:05:14 +0000 (19:05 +0000)]
Bump __FreeBSD_version for struct sysvec (sv_schedtail) changes.

13 years agoExtend struct sysvec with new method sv_schedtail, which is used for an
Dmitry Chagin [Tue, 8 Mar 2011 19:01:45 +0000 (19:01 +0000)]
Extend struct sysvec with new method sv_schedtail, which is used for an
explicit process at fork trampoline path instead of eventhadler(schedtail)
invocation for each child process.

Remove eventhandler(schedtail) code and change linux ABI to use newly added
sysvec method.

While here replace explicit comparing of module sysentvec structure with the
newly created process sysentvec to detect the linux ABI.

Discussed with: kib

MFC after: 2 Week

13 years agoCorrect readdir over ZFS handling.
Pawel Jakub Dawidek [Tue, 8 Mar 2011 18:39:41 +0000 (18:39 +0000)]
Correct readdir over ZFS handling.

Reported by: Pierre Beyssac <pb@fasterix.frmug.org>
MFC after: 1 month

13 years agoSome linux distros put mount point into the ext2fs labels, such as '/', or
Maxim Sobolev [Tue, 8 Mar 2011 17:00:31 +0000 (17:00 +0000)]
Some linux distros put mount point into the ext2fs labels, such as '/', or
'/boot', which confuses the devfs code and can cause userland programs to
fail reading /dev/ext2fs directory with weird error code, such as any
program that uses pwlib.

Strip any leading slashes before feeding the label to the geom_label code.

Sponsored by: Sippy Software, Inc.

MFC after: 1 week

13 years agoWhen constructing a new cpuset, apply the parent cpuset's mask to the new
John Baldwin [Tue, 8 Mar 2011 14:18:21 +0000 (14:18 +0000)]
When constructing a new cpuset, apply the parent cpuset's mask to the new
set's mask rather than the root mask.  This was causing the root mask to
be modified incorrectly.

Reviewed by: jeff
MFC after: 1 week

13 years agoUpdate to keep in sync with the HAL
Adrian Chadd [Tue, 8 Mar 2011 12:08:23 +0000 (12:08 +0000)]
Update to keep in sync with the HAL

13 years agoTunes and fixes the new DC-CC to seem to hit the
Randall Stewart [Tue, 8 Mar 2011 11:58:25 +0000 (11:58 +0000)]
Tunes and fixes the new DC-CC to seem to hit the
right mix.  Still may need some tweaks but it
appears to almost not give away too much to an
RFC2581 flow, but can really minimize the amount of
buffers used in the net.

MFC after: 3 months

13 years agoDo not assert buffer lock in VFS_STRATEGY() when kernel already paniced.
Konstantin Belousov [Tue, 8 Mar 2011 11:50:59 +0000 (11:50 +0000)]
Do not assert buffer lock in VFS_STRATEGY() when kernel already paniced.

Sponsored by: The FreeBSD Foundation
MFC after: 1 week

13 years ago- Bugfix: Root HUBs do not support re-enumeration.
Hans Petter Selasky [Tue, 8 Mar 2011 08:02:39 +0000 (08:02 +0000)]
- Bugfix: Root HUBs do not support re-enumeration.

MFC after: 14 days
Approved by: thompsa (mentor)

13 years agoBreak out the ath regulatory domain structures into a separate header file.
Adrian Chadd [Tue, 8 Mar 2011 07:42:09 +0000 (07:42 +0000)]
Break out the ath regulatory domain structures into a separate header file.

13 years agoImplement open-loop TX power control (OLC) for Merlin (AR9280) and
Adrian Chadd [Tue, 8 Mar 2011 06:59:59 +0000 (06:59 +0000)]
Implement open-loop TX power control (OLC) for Merlin (AR9280) and
generally tidy up the TX power programming code.

Enforce that the TX power offset for Merlin is -5 dBm, rather than
any other value programmable in the EEPROM. This requires some
further code to be ported over from ath9k, so until that is done
and tested, fail to attach NICs whose TX power offset isn't -5
dBm.

This improves both legacy and HT transmission on my merlin board.
It allows for stable MCS TX up to MCS15.

Specifics:

* Refactor out a bunch of the TX power calibration code -
  setting/obtaining the power detector / gain boundaries,
  programming the PDADC
* Take the -5 dBm TX power offset into account on Merlin -
  "0" in the per-rate TX power register means -5 dBm, not
  0 dBm
* When doing OLC
* Enforce min (0) and max (AR5416_MAX_RATE_POWER) when fiddling
  with the TX power, to avoid the TX power values from wrapping
  when low.
* Implement the 1 dBm cck power offset when doing OLC
* Implement temperature compensation for 2.4ghz mode when doing OLC
* Implement an AR9280 specific TX power calibration routine which
  includes the OLC twiddles, leaving the earlier chipset path
  (AR5416, AR9160) alone

Whilst here, use these refactored routines for the AR9285 TX power
calibration/programming code and enforce correct overflow/underflow
handling when fiddling with TX power values.

Obtained from: linux ath9k

13 years agocxgbe shouldn't directly know of the UMA zones where network buffers
Navdeep Parhar [Tue, 8 Mar 2011 03:04:07 +0000 (03:04 +0000)]
cxgbe shouldn't directly know of the UMA zones where network buffers
come from.

MFC after: 1 week

13 years agoOnly validate the partition setup when the user presses "Save". If the user
Nathan Whitehorn [Tue, 8 Mar 2011 01:17:30 +0000 (01:17 +0000)]
Only validate the partition setup when the user presses "Save". If the user
wants to exit without saving, it's not a problem if the disk is set up
in an invalid way.

13 years agosh: Test that . /dev/null returns exit status 0 and does not preserve $?.
Jilles Tjoelker [Mon, 7 Mar 2011 23:52:23 +0000 (23:52 +0000)]
sh: Test that . /dev/null returns exit status 0 and does not preserve $?.

Preserving $? may cause problems particularly if set -e is in effect.

It may be useful to preserve the old value of $? in the dot script but this
must not be implemented in such a way that it would break this test.

13 years agoSimplify uses of the web of pointers.
Konstantin Belousov [Mon, 7 Mar 2011 22:36:11 +0000 (22:36 +0000)]
Simplify uses of the web of pointers.

Reviewed by: mckusick
MFC after: 1 week

13 years ago- Remove superfluous ucom device lines in USB manual pages.
Hans Petter Selasky [Mon, 7 Mar 2011 21:53:15 +0000 (21:53 +0000)]
- Remove superfluous ucom device lines in USB manual pages.

Submitted by: Carl @ telus.net
MFC after: 14 days
Approved by: thompsa (mentor)

13 years agoRemove dead code that snuck in from the 32-bit PowerPC version of this
Nathan Whitehorn [Mon, 7 Mar 2011 20:32:46 +0000 (20:32 +0000)]
Remove dead code that snuck in from the 32-bit PowerPC version of this
file. The error being checked for does not exist on 64-bit systems.

13 years agoUnbreak the build.
Pawel Jakub Dawidek [Mon, 7 Mar 2011 19:54:51 +0000 (19:54 +0000)]
Unbreak the build.

MFC after: 2 weeks

13 years agoThe UFS dirhash code was attempting to update shared state in the dirhash
John Baldwin [Mon, 7 Mar 2011 18:33:29 +0000 (18:33 +0000)]
The UFS dirhash code was attempting to update shared state in the dirhash
from multiple threads while holding a shared lock during a lookup operation.
This could result in incorrect ENOENT failures which could then be
permanently stored in the name cache.

Specifically, the dirhash code optimizes the case that a single thread is
walking a directory sequentially opening (or stat'ing) each file.  It uses
state in the dirhash structure to determine if a given lookup is using the
optimization.  If the optimization fails, it disables it and restarts the
lookup.  The problem arises when two threads both attempt the optimization
and fail.  The first thread will restart the loop, but the second thread
will incorrectly think that it did not try the optimization and will only
examine a subset of the directory entires in its hash chain.  As a result,
it may fail to find its directory entry and incorrectly fail with ENOENT.

To make this safe for use with shared locks, simplify the state stored in
the dirhash and move some of the state (the part that determines if the
current thread is trying the optimization) into a local variable.  One
result is that we will now try the optimization more often.  We still
update the value under the shared lock, but it is a single atomic store
similar to i_diroff that is stored in UFS directory i-nodes for the
non-dirhash lookup.

Reviewed by: kib
MFC after: 1 week

13 years agoAdd missing declarations that I intended to commit with r219359.
David Schultz [Mon, 7 Mar 2011 16:05:45 +0000 (16:05 +0000)]
Add missing declarations that I intended to commit with r219359.
Not sure why we have math.h in lib/msun/ and complex.h in include/.

13 years agoCorrect a typo in the malloc(3) manpage. Malloc options are set in the
Ryan Stone [Mon, 7 Mar 2011 15:19:17 +0000 (15:19 +0000)]
Correct a typo in the malloc(3) manpage.  Malloc options are set in the
MALLOC_OPTIONS environment variable, not JEMALLOC_OPTIONS.

Reviewed by: jasone
Approved by: emaste (mentor)

13 years agoAdd AMD Geode CPU type to bsd.cpu.mk and examples/etc/make.conf
Martin Matuska [Mon, 7 Mar 2011 14:58:23 +0000 (14:58 +0000)]
Add AMD Geode CPU type to bsd.cpu.mk and examples/etc/make.conf
For CPUTYPE=core2 use -march=core2

PR: gnu/155308
MFC after: 2 weeks

13 years agoBackport Intel Core 2 and AMD Geode CPU types from gcc-4.3 (GPLv2)
Martin Matuska [Mon, 7 Mar 2011 14:48:22 +0000 (14:48 +0000)]
Backport Intel Core 2 and AMD Geode CPU types from gcc-4.3 (GPLv2)
These options are supported in this shape in all newer GCC versions.

PR: gnu/155308
Obtained from: gcc 4.3 (rev. 118090, 118973, 120846; GPLv2)
MFC after: 2 weeks

13 years agoPrint some of the numbers in human readable form (using %N).
Pawel Jakub Dawidek [Mon, 7 Mar 2011 10:42:28 +0000 (10:42 +0000)]
Print some of the numbers in human readable form (using %N).

MFC after: 2 weeks

13 years ago- Log size of data to synchronize in human readable form (using %N).
Pawel Jakub Dawidek [Mon, 7 Mar 2011 10:41:12 +0000 (10:41 +0000)]
- Log size of data to synchronize in human readable form (using %N).
- Log synchronization time (using %T).
- Log synchronization speed in human readable form (using %N).

MFC after: 2 weeks

13 years agoUse %S to print IP address and port number.
Pawel Jakub Dawidek [Mon, 7 Mar 2011 10:39:26 +0000 (10:39 +0000)]
Use %S to print IP address and port number.

MFC after: 2 weeks

13 years ago- Turn on printf extentions.
Pawel Jakub Dawidek [Mon, 7 Mar 2011 10:38:18 +0000 (10:38 +0000)]
- Turn on printf extentions.
- Load support for %T for pritning time.
- Add support for %N for printing number in human readable form.
- Add support for %S for printing sockaddr structure (currently only AF_INET
  family is supported, as this is all we need in HAST).
- Disable gcc compile-time format checking as this will no longer work.

MFC after: 2 weeks

13 years agoProvides three states for pjdlog_initialized, so we can also tell that
Pawel Jakub Dawidek [Mon, 7 Mar 2011 10:33:52 +0000 (10:33 +0000)]
Provides three states for pjdlog_initialized, so we can also tell that
this is fist initialization ever.

MFC after: 2 weeks

13 years agoTo be able to use printf extensions we need to turn off gcc format checking.
Pawel Jakub Dawidek [Mon, 7 Mar 2011 09:28:21 +0000 (09:28 +0000)]
To be able to use printf extensions we need to turn off gcc format checking.
Following the convention of NO_WERROR and NO_WCAST_ALIGN add NO_WFORMAT,
which, when defined in Makefile, turns off compile-time format checking
(by adding -Wno-format), but still allows to use high WARNS level.

MFC after: 2 weeks

13 years agoAdd cexp() to the complex(3) manpage. Thanks to bde for pointing out
David Schultz [Mon, 7 Mar 2011 08:54:20 +0000 (08:54 +0000)]
Add cexp() to the complex(3) manpage.  Thanks to bde for pointing out
that I missed this.

13 years agoRemove part of an uncommitted change that snuck into the last commit.
David Schultz [Mon, 7 Mar 2011 08:46:14 +0000 (08:46 +0000)]
Remove part of an uncommitted change that snuck into the last commit.

13 years agoRemove dead code.
Dmitry Chagin [Mon, 7 Mar 2011 08:12:07 +0000 (08:12 +0000)]
Remove dead code.

MFC after: 1 Week

13 years agoRemove unnecessary cast.
Stefan Farfeleder [Mon, 7 Mar 2011 07:31:15 +0000 (07:31 +0000)]
Remove unnecessary cast.

Reviewed by: jilles

13 years agoAdd some tests for cexp() and cexpf(). (I need to clean up all of
David Schultz [Mon, 7 Mar 2011 03:15:49 +0000 (03:15 +0000)]
Add some tests for cexp() and cexpf().  (I need to clean up all of
these tests some day, but in the mean time, they're a useful sanity
check for future changes.)

13 years agoConvert log10f() to use __kernel_log(), which is more accurate and simpler.
David Schultz [Mon, 7 Mar 2011 03:12:08 +0000 (03:12 +0000)]
Convert log10f() to use __kernel_log(), which is more accurate and simpler.

13 years agoConvert log10() to use __kernel_log(), which is more accurate and simpler.
David Schultz [Mon, 7 Mar 2011 03:11:27 +0000 (03:11 +0000)]
Convert log10() to use __kernel_log(), which is more accurate and simpler.

13 years agoAdd cexp() and cexpf().
David Schultz [Mon, 7 Mar 2011 03:09:24 +0000 (03:09 +0000)]
Add cexp() and cexpf().

Reviewed by: bde (earlier version)

13 years agoAdd support for printing out the open-loop TX power control EEPROM fields.
Adrian Chadd [Sun, 6 Mar 2011 23:15:24 +0000 (23:15 +0000)]
Add support for printing out the open-loop TX power control EEPROM fields.

13 years agoAllow to compress on-the-wire data using two algorithms:
Pawel Jakub Dawidek [Sun, 6 Mar 2011 23:09:33 +0000 (23:09 +0000)]
Allow to compress on-the-wire data using two algorithms:
- HOLE - it simply turns all-zero blocks into few bytes header;
it is extremely fast, so it is turned on by default;
it is mostly intended to speed up initial synchronization
where we expect many zeros;
- LZF - very fast algorithm by Marc Alexander Lehmann, which shows
very decent compression ratio and has BSD license.

MFC after: 2 weeks

13 years agombone is no longer a physical category
Doug Barton [Sun, 6 Mar 2011 23:01:02 +0000 (23:01 +0000)]
mbone is no longer a physical category

Submitted by: pav

13 years agoThe execution of the shebang script requires putting interpreter path,
Konstantin Belousov [Sun, 6 Mar 2011 22:59:30 +0000 (22:59 +0000)]
The execution of the shebang script requires putting interpreter path,
possible option and script path in the place of argv[0] supplied to
execve(2).  It is possible and valid for the substitution to be shorter
then the argv[0].

Avoid signed underflow in this case.

Submitted by: Devon H. O'Dell <devon.odell gmail com>
PR: kern/155321
MFC after: 1 week

13 years agoAllow to checksum on-the-wire data using either CRC32 or SHA256.
Pawel Jakub Dawidek [Sun, 6 Mar 2011 22:56:14 +0000 (22:56 +0000)]
Allow to checksum on-the-wire data using either CRC32 or SHA256.

MFC after: 2 weeks

13 years agosh(1): Reduce excessive semicolon-separated sentences.
Jilles Tjoelker [Sun, 6 Mar 2011 21:20:53 +0000 (21:20 +0000)]
sh(1): Reduce excessive semicolon-separated sentences.

Reported by: Benjamin Kaduk

13 years agoMake this makefile a little more make-like (e.g. build only things that
Nathan Whitehorn [Sun, 6 Mar 2011 21:19:33 +0000 (21:19 +0000)]
Make this makefile a little more make-like (e.g. build only things that
need to be rebuilt) and add an FTP-area target. Next will be docs and
memstick builds.

13 years agoTemporarily revert r219272; it breaks acl_is_trivial_np(3).
Edward Tomasz Napierala [Sun, 6 Mar 2011 20:12:09 +0000 (20:12 +0000)]
Temporarily revert r219272; it breaks acl_is_trivial_np(3).

13 years agoSimplify various code that allowed for sys_signame being lower case.
Jilles Tjoelker [Sun, 6 Mar 2011 19:50:47 +0000 (19:50 +0000)]
Simplify various code that allowed for sys_signame being lower case.

This was changed in r218285.

13 years agoBecause we call __printf_out() with a on-stack buffer, also call
Pawel Jakub Dawidek [Sun, 6 Mar 2011 19:47:46 +0000 (19:47 +0000)]
Because we call __printf_out() with a on-stack buffer, also call
__printf_flush() so we are sure it won't be referenced after we return.

MFC after: 2 weeks

13 years agoexpand_number() needs uint64_t, declare it here if not already declared.
Pawel Jakub Dawidek [Sun, 6 Mar 2011 17:46:06 +0000 (17:46 +0000)]
expand_number() needs uint64_t, declare it here if not already declared.

MFC after: 3 days

13 years agoInclude stdio.h, so we can include printf.h in any order, as it needs FILE.
Pawel Jakub Dawidek [Sun, 6 Mar 2011 17:45:37 +0000 (17:45 +0000)]
Include stdio.h, so we can include printf.h in any order, as it needs FILE.

MFC after: 2 weeks

13 years agoFix various issues in how %#T is handled:
Pawel Jakub Dawidek [Sun, 6 Mar 2011 17:43:32 +0000 (17:43 +0000)]
Fix various issues in how %#T is handled:
- If precision is 0, don't print period followed by no digits.
- If precision is 0 stop printing units as soon as possible
  (eg. if we have three years and five days and precision is 0
   print only 3y5d).
- If precision is not 0, print all units (eg. 3y0d0h0m0s.00).

MFC after: 2 weeks

13 years agoAdd some more IDs of HighPoint RocketRAID 64x.
Alexander Motin [Sun, 6 Mar 2011 16:10:39 +0000 (16:10 +0000)]
Add some more IDs of HighPoint RocketRAID 64x.

13 years ago- With the addition of TLS support binutils started to make the addend
Marius Strobl [Sun, 6 Mar 2011 15:20:11 +0000 (15:20 +0000)]
- With the addition of TLS support binutils started to make the addend
  values for resolved symbols relative to relocbase instead of sections
  so detect this case and handle as appropriate, which allows using
  kernel modules linked with affected versions of binutils. Actually I
  think this is a bug in binutils but given that apparently nobody
  complained for nearly six years and powerpc has basically the same
  workaround I decided to put it in for the sparc64 kernel, too.
- Fix R_SPARC_HIX22 relocations. Apparently these are hardly ever used.

13 years ago- Consistently abbreviate the names of the relocations.
Marius Strobl [Sun, 6 Mar 2011 13:25:46 +0000 (13:25 +0000)]
- Consistently abbreviate the names of the relocations.
- End sentences with dots.
- Fix whitespace.

13 years agoAdd missing bus_dmamap_sync() calls for the work DMA map.
Marius Strobl [Sun, 6 Mar 2011 13:08:25 +0000 (13:08 +0000)]
Add missing bus_dmamap_sync() calls for the work DMA map.

MFC after: 2 weeks

13 years agoAdd missing bus_dmamap_sync() calls for the work DMA map.
Marius Strobl [Sun, 6 Mar 2011 13:06:41 +0000 (13:06 +0000)]
Add missing bus_dmamap_sync() calls for the work DMA map.

MFC after: 2 weeks

13 years ago- Allocate the DMA memory used for the work area as coherent as at least
Marius Strobl [Sun, 6 Mar 2011 12:54:00 +0000 (12:54 +0000)]
- Allocate the DMA memory used for the work area as coherent as at least
  the ataahci(4) and atamarvell(4) drivers share it between the host and
  the controller.
- Spell some zeros as BUS_DMA_WAITOK when used as bus_dmamem_alloc() flags.

MFC after: 2 weeks

13 years ago- Allocate the DMA memory shared between the host and the controller as
Marius Strobl [Sun, 6 Mar 2011 12:48:15 +0000 (12:48 +0000)]
- Allocate the DMA memory shared between the host and the controller as
  coherent.
- Add some missing bus_dmamap_sync() calls. This includes putting such
  calls before calling reply handlers instead of calling bus_dmamap_sync()
  for the request queue from individual reply handlers as these handlers
  generally read back updates by the controller.

Tested on amd64 and sparc64.

MFC after: 2 weeks

13 years agoMention setloginclass(2) in login_class(3).
Edward Tomasz Napierala [Sun, 6 Mar 2011 08:56:44 +0000 (08:56 +0000)]
Mention setloginclass(2) in login_class(3).

13 years agoMove getloginclass(2) and setloginclass(2) to FBSD_1.2, where they should've
Edward Tomasz Napierala [Sun, 6 Mar 2011 08:55:36 +0000 (08:55 +0000)]
Move getloginclass(2) and setloginclass(2) to FBSD_1.2, where they should've
been added in the first place.

13 years agoAdd FBSD_1.2; syscalls added in 9-CURRENT are supposed to go there.
Edward Tomasz Napierala [Sun, 6 Mar 2011 08:52:59 +0000 (08:52 +0000)]
Add FBSD_1.2; syscalls added in 9-CURRENT are supposed to go there.

Suggested by: kib

13 years agoAdd manual page for getloginclass(2) and setloginclass(2).
Edward Tomasz Napierala [Sun, 6 Mar 2011 08:35:50 +0000 (08:35 +0000)]
Add manual page for getloginclass(2) and setloginclass(2).

13 years agoFix libzpool build.
Pawel Jakub Dawidek [Sun, 6 Mar 2011 01:22:14 +0000 (01:22 +0000)]
Fix libzpool build.

MFC after: 1 month

13 years agoObligatory first commit to the include myself in the list of
Steve Kargl [Sun, 6 Mar 2011 00:57:08 +0000 (00:57 +0000)]
Obligatory first commit to the include myself in the list of
committers list.

Approved by: das

13 years agoAdd an EEPROM op that extracts out the power table offset.
Adrian Chadd [Sun, 6 Mar 2011 00:30:43 +0000 (00:30 +0000)]
Add an EEPROM op that extracts out the power table offset.
It defaults to -5 dBm for eeproms earlier than v21.

This apparently only applies to Merlin (AR9280) or later,
earlier 11n chipsets have a power table offset of 0.
All the code in ath9k which checks the power table offset
and takes it into account first ensures the chip is
Merlin or later.

13 years agoMake renaming of a ZVOL, ZVOL's parent directory and ZVOL snapshot work.
Pawel Jakub Dawidek [Sat, 5 Mar 2011 22:31:03 +0000 (22:31 +0000)]
Make renaming of a ZVOL, ZVOL's parent directory and ZVOL snapshot work.

Reported by: avg
MFC after: 1 month

13 years agoSimplify zvol_remove_minors() a bit.
Pawel Jakub Dawidek [Sat, 5 Mar 2011 22:24:31 +0000 (22:24 +0000)]
Simplify zvol_remove_minors() a bit.

MFC after: 1 month

13 years agoChange HALDEBUG() to be a macro that conditionally calls the debug output routine.
Adrian Chadd [Sat, 5 Mar 2011 21:20:18 +0000 (21:20 +0000)]
Change HALDEBUG() to be a macro that conditionally calls the debug output routine.

The earlier way of doing debugging would evaluate the function parameters
before calling the HALDEBUG. In the case of detailed register debugging
would mean a -lot- of unneeded register IO and other stuff was going on.

This method evaluates the ath_hal_debug variable before the function
parameters are evaluated, drastically reducing the amount of overhead
enabling HAL debugging during compilation.

13 years agoStyle(9) fix.
Dmitry Chagin [Sat, 5 Mar 2011 20:54:17 +0000 (20:54 +0000)]
Style(9) fix.
Fix indentation in comment, double ';' in variable declaration.

MFC after: 1 Week

13 years agoPartially reworked r219042.
Dmitry Chagin [Sat, 5 Mar 2011 20:36:42 +0000 (20:36 +0000)]
Partially reworked r219042.
The reason for this is a bug at ktrops() where process dereferenced
without having a lock. This might cause a panic if ktrace was runned
with -p flag and the specified process exited between the dropping
a lock and writing sv_flags.

Since it is impossible to acquire sx lock while holding mtx switch
to use asynchronous enqueuerequest() instead of writerequest().

Rename ktr_getrequest_ne() to more understandable name [1].

Requested by: jhb [1]

MFC after: 1 Week

13 years agoAdd two missing vertical bars.
Gavin Atkinson [Sat, 5 Mar 2011 15:18:11 +0000 (15:18 +0000)]
Add two missing vertical bars.

13 years agoExport login class information via kinfo and make it possible to view
Edward Tomasz Napierala [Sat, 5 Mar 2011 14:41:49 +0000 (14:41 +0000)]
Export login class information via kinfo and make it possible to view
it using "ps -o class".

13 years agosh: Fix some warnings in code for arithmetic expressions.
Jilles Tjoelker [Sat, 5 Mar 2011 13:27:13 +0000 (13:27 +0000)]
sh: Fix some warnings in code for arithmetic expressions.

Submitted by: eadler

13 years agoRegenerate.
Edward Tomasz Napierala [Sat, 5 Mar 2011 12:46:24 +0000 (12:46 +0000)]
Regenerate.

13 years agoAdd two new system calls, setloginclass(2) and getloginclass(2). This makes
Edward Tomasz Napierala [Sat, 5 Mar 2011 12:40:35 +0000 (12:40 +0000)]
Add two new system calls, setloginclass(2) and getloginclass(2).  This makes
it possible for the kernel to track login class the process is assigned to,
which is required for RCTL.  This change also make setusercontext(3) call
setloginclass(2) and makes it possible to retrieve current login class using
id(1).

Reviewed by: kib (as part of a larger patch)

13 years agoBe sure to stay within the bounds of the mod_str array when displaying
Navdeep Parhar [Sat, 5 Mar 2011 04:19:38 +0000 (04:19 +0000)]
Be sure to stay within the bounds of the mod_str array when displaying
the transceiver type.

13 years agoThere is no need to hold an ingress queue's lock while processing its
Navdeep Parhar [Sat, 5 Mar 2011 04:04:23 +0000 (04:04 +0000)]
There is no need to hold an ingress queue's lock while processing its
descriptors.

MFC after: 1 week

13 years agoCalculate how many descriptors can be reclaimed before calling
Navdeep Parhar [Sat, 5 Mar 2011 03:54:37 +0000 (03:54 +0000)]
Calculate how many descriptors can be reclaimed before calling
reclaim_tx_descs

13 years agoTweaks for rx:
Navdeep Parhar [Sat, 5 Mar 2011 03:42:03 +0000 (03:42 +0000)]
Tweaks for rx:

- everything related to LRO should be in #ifdef INET blocks
- reorder sge_iq's fields so that the most frequently used are all together
- pull all rx code into t4_intr_data directly
- let go of the ingress queue lock when passing up data
- refill the freelist only if it is short of at least 32 buffers

13 years agoStore the ifnet rather than the port_info in each txq and rxq struct.
Navdeep Parhar [Sat, 5 Mar 2011 03:27:14 +0000 (03:27 +0000)]
Store the ifnet rather than the port_info in each txq and rxq struct.

MFC after: 1 week

13 years agoA txpkts work request should have a valid FID.
Navdeep Parhar [Sat, 5 Mar 2011 03:18:56 +0000 (03:18 +0000)]
A txpkts work request should have a valid FID.

MFC after: 1 week

13 years agoUpgrade the firmware on the card automatically if a better version is
Navdeep Parhar [Sat, 5 Mar 2011 03:12:50 +0000 (03:12 +0000)]
Upgrade the firmware on the card automatically if a better version is
available.  Downgrade only for a major version mismatch.

MFC after: 1 week

13 years agoResume tx immediately in response to an SGE egress update from the hardware.
Navdeep Parhar [Sat, 5 Mar 2011 03:06:38 +0000 (03:06 +0000)]
Resume tx immediately in response to an SGE egress update from the hardware.

MFC after: 1 week

13 years agoFix incorrect assertion.
Navdeep Parhar [Sat, 5 Mar 2011 03:01:14 +0000 (03:01 +0000)]
Fix incorrect assertion.

MFC after: 3 days

13 years agoFlush both reads *and* writes to registers.
Matt Jacob [Sat, 5 Mar 2011 00:59:34 +0000 (00:59 +0000)]
Flush both reads *and* writes to registers.

Obtained from: Miod Vallat in OpenBSD
MFC after: 1 week

13 years agoUse ffs() to locate free bits in the inode bitmap rather than a loop with
John Baldwin [Fri, 4 Mar 2011 22:26:41 +0000 (22:26 +0000)]
Use ffs() to locate free bits in the inode bitmap rather than a loop with
bit shifts.

Reviewed by: mckusick
MFC after: 1 month

13 years agoFix a panic that can happen when trying to destroy a lagg(4) with scheduler set to...
Ermal Luçi [Fri, 4 Mar 2011 20:37:38 +0000 (20:37 +0000)]
Fix a panic that can happen when trying to destroy a lagg(4) with scheduler set to none.

Approved by: thompsa(mentor)
MFC after: 1 week

13 years agoMake UFS use PSARC/2010/029 NFSv4 ACL semantics by default, just like
Edward Tomasz Napierala [Fri, 4 Mar 2011 19:53:07 +0000 (19:53 +0000)]
Make UFS use PSARC/2010/029 NFSv4 ACL semantics by default, just like
ZFSv28 does.

MFC after: 2 months

13 years agoPOSIX.1-2008 moved some constants from the XSI option to the Base.
Jilles Tjoelker [Fri, 4 Mar 2011 19:28:27 +0000 (19:28 +0000)]
POSIX.1-2008 moved some constants from the XSI option to the Base.

MFC after: 2 weeks

13 years agoOne more fix. Now all ACL tests pass again.
Edward Tomasz Napierala [Fri, 4 Mar 2011 18:46:19 +0000 (18:46 +0000)]
One more fix.  Now all ACL tests pass again.

13 years agoAdapt tools-crossfs.test to the new semantics.
Edward Tomasz Napierala [Fri, 4 Mar 2011 18:37:26 +0000 (18:37 +0000)]
Adapt tools-crossfs.test to the new semantics.

13 years agoMake acl_strip_np(3) use new trivial ACL format for NFSv4 ACls (three
Edward Tomasz Napierala [Fri, 4 Mar 2011 17:23:10 +0000 (17:23 +0000)]
Make acl_strip_np(3) use new trivial ACL format for NFSv4 ACls (three
entries instead of six).  This makes "setfacl -b" do the right thing
for ACLs on ZFS.  UFS recognizes both kinds of trivial ACLs; no change
there.

MFC after: 2 months

13 years agoAdapt NFSv4 ACL regression test to the fact that the new ZFS uses
Edward Tomasz Napierala [Fri, 4 Mar 2011 17:10:39 +0000 (17:10 +0000)]
Adapt NFSv4 ACL regression test to the fact that the new ZFS uses
new semantics.

13 years agoMake ACL tests slightly easier to run.
Edward Tomasz Napierala [Fri, 4 Mar 2011 17:07:02 +0000 (17:07 +0000)]
Make ACL tests slightly easier to run.

13 years ago- Add a FEATURE for capsicum (security_capabilities).
Alexander Leidinger [Fri, 4 Mar 2011 09:03:54 +0000 (09:03 +0000)]
- Add a FEATURE for capsicum (security_capabilities).
- Rename mac FEATURE to security_mac.

Discussed with: rwatson

13 years agoAdd the Buffalo (Melco Inc.) WLI-UC-G301N
Daichi GOTO [Fri, 4 Mar 2011 07:01:45 +0000 (07:01 +0000)]
Add the Buffalo (Melco Inc.) WLI-UC-G301N

PR: usb/155229
Submitted by: Yoshiaki UCHIKAWA
MFC after: 1 week

13 years agoFix some _POSIX minimum/maximum values in limits.h:
Jilles Tjoelker [Thu, 3 Mar 2011 22:34:13 +0000 (22:34 +0000)]
Fix some _POSIX minimum/maximum values in limits.h:
* Some values changed in POSIX.1-2001; provide the former value if a program
  requests compliance to an earlier version of POSIX. [1]
* Add missing _POSIX_CLOCKRES_MIN constant. This is a maximum value but
  otherwise works the same as the minimum values.

PR: standards/104743
Submitted by: bde [1] (not exact #ifdefs, but the values)
MFC after: 2 weeks